The algorithm presented so far is woefully inefficient for pedestrian graphs. Consider rendering
.
Rendering 
with
pixel-based testing would require more than
one million interval evaluations of S.
Efficiency may be improved considerably by using
super-pixel testing: testing a group of pixels
with a single interval evaluation of S.
Automated symbolic reasoning may deduce that
with
column-based testing would require more than
one thousand interval evaluations of S.
Efficiency may be improved considerably by using
super-column testing: testing a group of pixels
with a single interval evaluation of S.
With either method, partial information presented during rendering may be of limited utility. Super-pixel, and super-column, testing may present informative intermediate renderings.
| Jeff Tupper | March 1996 |